Combining Observational and Randomized Data for Estimating Heterogeneous Treatment Effects

Hatt, Tobias; Berrevoets, Jeroen; Curth, Alicia; Feuerriegel, Stefan; Schaar, Mihaela van der

doi:10.48550/arxiv.2202.12891

Cited by 7 publications

(7 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, when patients' information comes from both randomized trials and real-world data, there are two main approaches for integrative analysis: meta analysis and pooled patient data analysis. See the related recent work by Hatt et al [2022], Yang et al [2020], Cheng and Cai [2021] and Athey et al [2020].…”

Section: Precision Medicine and Dynamic Treatment Regimesmentioning

confidence: 99%

Dynamic treatment regimes with interference

2022

View full text Add to dashboard Cite

Precision medicine describes health care where patient‐level data are used to inform treatment decisions. Within this framework, dynamic treatment regimes (DTRs) are sequences of decision rules that take individual patient information as input data and then output treatment recommendations. DTR estimation from observational data typically relies on the assumption of no interference: i.e., the outcome of one individual is unaffected by the treatment assignment of others. However, in many social network contexts, such as friendship or family networks, and for many health concerns, such as infectious diseases, this assumption is questionable. We investigate the DTR estimation method of dynamic weighted ordinary least squares (dWOLS), which boasts of easy implementation and the so‐called double‐robustness property, but relies on the assumption of no interference. We define a network propensity function and build on it to establish an implementation of dWOLS that remains doubly robust under interference associated with network links. The method's properties are demonstrated via simulation and applied to data from the Population Assessment of Tobacco and Health (PATH) study to investigate cigarette dependence within two‐person household networks.

show abstract

Section: Precision Medicine and Dynamic Treatment Regimesmentioning

confidence: 99%

Dynamic treatment regimes with interference

2022

View full text Add to dashboard Cite

show abstract

“…Extensive works focus on treatment effect estimation in static settings [36,11,14]. Two important methods that adopt machine learning for average treatment effect estimation in the static setting are: (i) targeted maximum likelihood estimation (TMLE) [42] and (ii) DragonNet [37].…”

Section: Causal Effect Estimation In the Static Settingmentioning

confidence: 99%

Estimating average causal effects from patient trajectories

Frauen¹,

Hatt²,

Melnychuk³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

In medical practice, treatments are selected based on the expected causal effects on patient outcomes. Here, the gold standard for estimating causal effects are randomized controlled trials; however, such trials are costly and sometimes even unethical. Instead, medical practice is increasingly interested in estimating causal effects among patient subgroups from electronic health records, that is, observational data. In this paper, we aim at estimating the average causal effect (ACE) from observational data (patient trajectories) that are collected over time. For this, we propose DeepACE: an end-to-end deep learning model. DeepACE leverages the iterative Gcomputation formula to adjust for the bias induced by time-varying confounders. Moreover, we develop a novel sequential targeting procedure which ensures that DeepACE has favorable theoretical properties, i. e., is doubly robust and asymptotically efficient. To the best of our knowledge, this is the first work that proposes an end-to-end deep learning model for estimating time-varying ACEs. We compare DeepACE in an extensive number of experiments, confirming that it achieves state-of-the-art performance. We further provide a case study for patients suffering from low back pain to demonstrate that DeepACE generates important and meaningful findings for clinical practice. Our work enables medical practitioners to develop effective treatment recommendations tailored to patient subgroups.

show abstract

“…Another question is how the CATE estimators depend on the ratio of numbers of treatments and controls in the training set. We study the case when the number of controls c is 200 and the ratio takes values The next experiments allow us to investigate how the CATE estimators depend on the value of hyperparameter α which controls the impact of the control and treatment networks in the loss function (17).…”

Section: Experiments With Different Values Of the Treatment Ratiomentioning

confidence: 99%

“…Many methods for estimating CATE have been proposed and developed due to importance of the problem in medicine and other applied areas [10,11,12,13,14,15,16,17,18,19,20,21,22,23,24]. This is only a small part of all publications which are devoted to solving the problem of estimating CATE.…”

Section: Introductionmentioning

confidence: 99%

Heterogeneous Treatment Effect with Trained Kernels of the Nadaraya-Watson Regression

Konstantinov¹,

R.²,

Utkin³

2022

Preprint

View full text Add to dashboard Cite

A new method for estimating the conditional average treatment effect is proposed in the paper. It is called TNW-CATE (the Trainable Nadaraya-Watson regression for CATE) and based on the assumption that the number of controls is rather large whereas the number of treatments is small. TNW-CATE uses the Nadaraya-Watson regression for predicting outcomes of patients from the control and treatment groups. The main idea behind TNW-CATE is to train kernels of the Nadaraya-Watson regression by using a weight sharing neural network of a specific form. The network is trained on controls, and it replaces standard kernels with a set of neural subnetworks with shared parameters such that every subnetwork implements the trainable kernel, but the whole network implements the Nadaraya-Watson estimator. The network memorizes how the feature vectors are located in the feature space. The proposed approach is similar to the transfer learning when domains of source and target data are similar, but tasks are different. Various numerical simulation experiments illustrate TNW-CATE and compare it with the well-known T-learner, S-learner and X-learner for several types of the control and treatment outcome functions. The code of proposed algorithms implementing TNW-CATE is available in https://github.com/Stasychbr/TNW-CATE.

show abstract

Combining Observational and Randomized Data for Estimating Heterogeneous Treatment Effects

Cited by 7 publications

References 51 publications

Dynamic treatment regimes with interference

Dynamic treatment regimes with interference

Estimating average causal effects from patient trajectories

Heterogeneous Treatment Effect with Trained Kernels of the Nadaraya-Watson Regression

Contact Info

Product

Resources

About